Linear and helical flow in a perfluorosulfonate membrane of annular

Petr Kuban, Purnendu K. Dasgupta, and Kavin A. Morris .... M Cedillo , Sivakumar Jambunathan , Sofia G Hohnholt , Purnendu K Dasgupta , Duane K Wolcot...
0 downloads 0 Views 1MB Size
96 (9) (10) (11) (12) (13) (14)

(15) (16) (17) (18) (19) (20) (21)

Anal. Chem. 1904, 56,96-103 Treiber, L. R. J. Chromatogr. 1974, 100, 123. Hurtubise, R. J. Anal. Chem. 1977, 49, 2160. Goldman, J. J. Chromatogr. 1973, 78,7. Yamamoto, H.; Kurita, T.; Suzuki, J.; Hira, R.; Nakamo, K.; Makabe, H. J. Chromatogr. 1976, 116, 29. Bush, I.E. J. Chromatogr. 1967, 2 9 , 157. Bush, I. E. I n “Methods of Biochemical Analysis”; Glick, D., Ed.; Wiley: New York, 1963; Vol. ll, p 149. Duntley, S. Q. J. Opt. Soc. Am. 1942, 32,61. Lugg, J. W. H.J. Chromatogr. 1963, 10, 272. Spitzer, D.; Ten Bosch, J. J. Appl. Opt. 1976, 15, 934. Forresier, J. W. “Principles of Systems”; Wright Allen Press: Cambridge, MA, 1968; Chapter 8. Bush, I.E., unpublished results. Goidman, J., Imperial Chemlcal Industries, Pharmaceutical Division, Macclesfield, U.K., personal comrnunlcations. Keilin. D.; Hartree, E. F. Nature (London) 1949, 164, 254.

Pollak, V. Adv. Chromatogr. 1979, 17, 1. Goldman, J.; Goodail, R. R. J. Chromatogr. 1969, 40, 345. Goldman, J.; Goodall, R. R. J. Chromatogr. 1970, 47, 386. Goodail, R. R. J. Chromatogr. 1975, 103,265. Treiber, L. R.; Ortengren, 8.; Llndsten, R.; Ortengren, T. J. Chromatogr. 1972, 73, 135. (27) Touchstone, J. C.; Levin, S. S.; Murawec, T. Anal. Chem. 1071, 43, 858. (28) Knapstein, P.; Touchstone, J. C. J. Chromatogr. 1966, 37,83. (29) Boulton, A. A. I n “Methods in Biochemical Analysis”; Glick, D., Ed.; Wiley: New York, 1968; Vol. 16, p 327. (30) Rayieigh, Lord Proc. R. SOC. London, Ser. A 1914, 90,219. (22) (23) (24) (25) (26)

RECEIVED for review June 29, 1983. Accepted October 12, 1983.

Linear and Helical Flow in a Perfluorosulfonate Membrane of Annular Geometry as a Continuous Cation Exchanger Purnendu K. Dasgupta

Department of Chemistry, Texas Tech University, Box 4260, Lubbock, Texas 79409

Flow through an annulus, especlally helical flow through an annulus, permits highly efficient mass transfer to the walls of a tube and can be constructed in low dead volume conflguratlons. Mass transfer efficiency has been studied with K+ and NEt4+as dlffuslng species, as a function of annular gap and hellx diameter, In hellces of nylon monofilament filled Naflon perfluorosulfonate membrane tubing. Penetration of common acid regenerants through Naflon membrane, used for proton exchange, has been studled In detall and may be predlcted as a functlon of concentratlon. Mass transfer to, through, and from the wall has been consldered In detail. I n regard to mass transfer to the wall, annular hellcal flow was found to be hlghly efflcient.

Our original interest in continuous ion exchange stemmed from extending the concept of “diffusion denuding” of gases. Diffusion denuders are widely used in atmospheric analysis for discriminating between gaseous and aerosol species where a chemical discrimination is difficult ( I ) . When the mixture of a gas and particles (e.g., atmospheric NH3 and aerosols containing NH4+)is passed through a suitably long tube whose walls act like a perfect sink for the gas (e.g., coated with oxalic acid), the NH3 is selectively removed by diffusion to the wall because of the 4 orders of magnitude higher diffusion coefficient for the gas compared to a micrometer sized atmospheric particle (2). The classical equation that describes mass transfer to the walls of a cylindrical tube from a stream flowing through the tube is due to Gormley and Kennedy (3)and has been verified experimentally by a number of workers. The sole parameter governing mass transfer under laminar flow conditions is n D L / F where D is the diffusion coefficient of the species to be removed, L is the length of the tube, and F is the flow rate. Although diffusion coefficients of ions in the liquid phase are much smaller than gaseous diffusion coefficients, a comparable and compensating difference exists between air sampling and chromatographic flow rates. The concept of ion exchange by diffusion denuding (IEDD) with an ion exchanger tubing was then tested successfully with perfluorosulfonate Nafion tubing which were in use in this laboratory for the removal of water vapor by diffusion denuding. Similar to the use of dry airflow outside the tube to

maintain its sink capacity for use as a dryer ( 4 ) ,a flow of dilute acid was used outside the tubing for maintaining proton exchange capacity. A principal objective of such a continuous proton exchanger was utilization in ionic chromatography. Although quantitative IEDD was accomplished, we deemed the band dispersion in such hollow tubes (700 pm diameter) too large to be useful for chromatographic purposes. The Gormley-Kennedy equation states that the mass transfer efficiency is independent of tube diameter. In contrast, relative band dispersion is proportional to the fourth power of diameter (5). Thus, dispersion problems should be greatly reduced, without affecting ion exchange efficiency, by a reduction of diameter. At this time, we noted the prior and independent work of Stevens et al. (6), describing a “hollow fiber suppressor” which utilized 300 pm diameter sulfonated polyethylene tubing as the ion exchange membrane tube. This work also describes the detailed principle of operation of the continuous cation exchanger, which is therefore not repeated here. Interestingly, even though the tube diameter was lower by more than a factor of 2 compared to the tube diameter used by us, the band dispersion was greater than what we were observing. Apparently, this dispersion was due to the much longer tube lengths that were required. Any of the following may limit the overall exchange rate observed with the system described above: (a) mass transfer to the wall, (b) mass transfer through the wall, and (c) mass transfer from the wall (regenerant flow/concentration insufficient). Assuming that sufficient exchange sites were present and regenerant flow was adequate, the only possible explanation is limitation of exchange by (b). Mass transfer through the wall is rate limited by the diffusion coefficient of the ion through the wall matrix and the wall thickness. Since the wall thickness of the perfluorosulfonate tubing was nearly double that of the polyethylene tubing, ion transport through the hydrophilic perfluorosulfonate must be far more facile. Having failed to persuade the manufacturer toward making smaller diameter perfluorosulfonate tubing available (7), we sought alternatives for reducing band broadening problems with available perfluorosulfonate tubing. Theoretical and experimental studies on flow through an open tubular helix are not new (8-13). The helical structure leads to the development of a secondary flow perpendicular

0003-2700/84/0356-0096$01.50/00 1983 American Chemical Society

.

ANALYTICAL CHEMISTRY, VOL. 56, NO. 1, JANUARY 1984

t o the axial flow. This secondary flow flattens the parabolic profile of the flow velocity observed in a linear tube and also leads to both reduced axial dispersion and increased mass transfer to the walls of the tube. Following the pioneering work of Dean (8,9), the most important parameter in helical flow, Re(dt/d,)1/2, has been designated the Dean number (Re is Reynold's number and dt and d, are the diameters of the tube and the coil, respectively). T h e differences between helical and linear flow increase with increasing Dean number. The rigorous analytical solution for the hydrodynamics of helical flow is complex. Although numerical methods have recently become available (14,15),certain approximations are necessary and such methods apply to a limited, albeit useful, range of Dean numbers. For annular flow, mass transfer studies have not received much attention. However, the analogous case for heat transfer has been studied in detail (16)and predicts that transfer should be more efficient compared to flow through an open tube, the difference being significant for small values of ro/ri where r, and ri are the radii of the outer and inner cylinder, respectively. Very recently, mass transfer in an annular diffusion denuder has been studied both theoretically and experimentally (17). In this study, although both the surfaces defining the annulus act as sinks, a clear relationship between increased mass transfer efficiency and decreasing annular gap (low ro/ri) was established. Additionally, the advantages of small annular gap systems in promoting mass transport in boundary layers have been recognized for a long time in designing efficient dialyzers (18). During the course of our study, the work of Stevens et al. (19),which employed Nafion tubing packed with inert beads of optimized size, was published. We pursued our investigations because it was clear that helical flow through an annular geometry may be a significant improvement over the bead packed configuration. Comparison of the annular helical device with the packed bead device, as suppressors in anion chromatography, appears in a companion paper (20).

EXPERIMENTAL SECTION All reagents used in this work were of analytical reagent grade. The water used was distilled and deionized but not free from COz. Solutions to be ion exchanged were thermostated at 30 f 0.1 "C and pumped by an Altex Model llOA reciprocating pump (Berkeley, CA) through a pulse dampener (Handy and Herman Tube Co., Norristown, PA) into the device under test. The effluent conductivity was measured with a Wescan Model 213 conductivity detector (Santa Clara, CA). The attendant detector cells (nominal volume 2 or 20 pL, cell constants as determined with 0.01 D KC1 solution were 34.1 and 0.117, respectively) were thermally insulated by polyurethane foam molds. The reported conductance values are corrected to 25 "C. The regenerant solution, also thermostated at 30 "C, was pumped by an Eldex Model E-120-S reciprocating pump (Menlo Park, CA) or allowed to flow under gravity. Construction of Devices. Nafion 811x perfluorosulfonate membrane tubing (Du Pont Polymer Products Division, Wilminton, DE) was used as the cation exchanger membrane tubing. The nominal dimensions of this tubing are 875 pm 0.d. and 625 pm i.d. in the dry state. The hydrophilic material expands s i g nificantly upon wetting. Two separate batches of tubing were used during the course of this study; their diameters were perceptibly different although both were within the usual manufacturing tolerance. Devices otherwise identical, except for the batch of tubing, produced significantly different results. The bulk of the data reported in this paper are typical among several dozen individual devices fabricated with the smaller diameter tubing, which is at the lower end of the diameter distribution of the manufacturer's product (21). Options for leak-freeconnections (to withstand several hundred pounds per square inch) for membrane tubing with 75 pm wall thickness are limited. Because of wide interest in the potential application of Nafion membranes (22, 23), sealing techniques developed during the course of this work may be useful to others. i

C D E

F

97

G

(b)

Flgure 1. Inlet/outlet designs: A,K, inlet/outlet PTFE tube; B, 1/4-28 male nut; C, O-rings; D,O, microbore PTFE insert: E,N, Nafion membrane tubing; F, 1/4-28 threaded T: L, polypropylene T; G,P, nylon monofilament; M, nichrome wire crimp; J, H, adhesive filler; I, jacket tubing. The swell-seal method (19, 23) which takes advantage of the hydrophilic expansion of Nafion, could not be easily applied to filament filled Nafion tubing. Much initial work was done with a flange seal technique in which the open end of the tubing is brought close to a flame whereupon it softens, rolls back upon itself and produces a ring like flange. While with appropriately constructed fittings a good seal could be obtained, commercially available T-fittings with 1/4-28 thread (made for 1/16 in. tubing) frequently have interconnecting holes that are too large for this purpose. Also, very small homemade washers were needed for sealing. Although this technique has some distinct advantages, this approach was discontinued in favor of the following. A small section (0.5-1 cm) of microbore PTFE tubing (0.3 mm i.d., 0.75 mm 0.d.) was forced into the Nafion tubing and then the Nafion tubing was inserted into the connecting PTFE tubing (0.3 mm i.d., 1.5 mm 0.d.; the tubing was enlarged at the end with a suitable tool to accommodate the Nafion tubing; if dead volume considerations are unimportant, 0.8 mm id., 1.6 mm 0.d. PTFE tubing is more convenient) to no further depth than the microbore insert. Sealing was then accomplished either by compression with O-rings (Figure la) or by crimping with 28 gauge nichrome wire (Figure lb). Monofilament nylon lines (fishing lines) were used for insertion into the Nafion tubing. Three different filament diameters were investigated: 0.52 mm (Sigma, 25 lb strength line), 0.56 and 0.66 mm (Stren, Du Pont, 30 and 40 lb strength lines, respectively). If the membrane is thoroughly wet, insertion of the 0.52- or 0.56-mm filaments into the Nafion tubing is facile. The direct insertion of the 0.66-mm filament (which actually exceeds the dry inside diameter of the tubing) is however, very difficult. Initially we accomplished this by splicing the filament onto a narrower filament which has already been inserted into a desired length of the Nafion tubing and pulling it through. This was found to be very difficult for lengths exceeding 0.5 m and a better method was found. After insertion of the microbore PTFE tubing and establishment of a leak-free connection at one end, this end was connected to a pump and water was pumped through the Nafion tubing. The filament was pushed from the open end of the Ndion tubing inward. A flow rate of 1mL/min produced a pressure drop about 200 psi which adequately expanded the tubing for smooth passage of the filament. When the filament was all the way in and butting the microbore tubing, the tubing and filament were gripped from the insertion end and the pressure was released by discontinuing flow. Excess filament was then cut off flush with the tubing. Gentle stretching of the membrane tubing then produced enough void space at the insertion end for inserting a small section of microbore tubing for connections, as at the other end.

98

ANALYTICAL CHEMISTRY, VOL. 56, NO. 1, JANUARY 1984

--

Table I. Regenerant Penetrationa external liquid

H2O HCl, 1 mM HCIO,, 1 mM HNO,, 1 mM

/

conductance, PS

5.3 6.4 6.3 6.7

external liquid

conductance, pS

H,SO,, 0.5 mM

6.2 20.0 32.6 1400

H,PO,, 0.33 mM

HOAc, 1 mM HOAc, pH 3

a Internal liquid water, flow rate 1 mL/min. External flow rate 2 mL/min. Membrane tube length 20 cm. Detector cell 20 bL.

H

I

J

K

L

Cb)

Flgure 2. Device designs: A, 1/4-28 T; B, adhesive filler; C, support rod; D, enlarged cross section below; E, filament; F, Nafion tubing; 0, jacket tubing; H, adhesive; I, polypropylene T J, jacket tube; K, filament filled helix; L, enlarged cross section below; E, filament; F, Nafion tubing.

Compared to the original unstretched dry lengths, the longitudinal expansion of the Nafion tubing as a result of the above operations is significant and can be as large as 10%. The tubing lengths reported in this paper refer to the original dry lengths. Coiling was then accomplished to produce the filament filled helix (FFH). Two basic types of device construction were studied. In the first design, the jacket tubing for regenerant flow outside the membrane tubing was concentric with the membrane tubing and filament, i.e., the jacket formed an outer helix (Figure 2a). In this design, used with the inlet connection depicted in Figure la, jacket inside diameter only slightly larger than the membrane outside diameter were practical. Vinyl tubing (1-1.25 mm i.d.) was used as jacket and was sealed in place with a suitable adhesive. For larger diameter helices, 3 mm 0.d. PTFE tubing, which could be conveniently flange sealed, were used as jacket tubing. The support rod and the inlet/outlet T’s were cemented together. Although the concentric design permits the most efficient use of the regenerant flow, the attainable lower limit of helix diameter was poor. Additionally, attempts to obtain low diameter helices with small diameter support rods (e.g., 1.5 mm 0.d. metal tubing) were frequently foiled by the membrane tubing biting into the jacket and effectively losing the surface area of the membrane in such contact from an active role. The advantage of this design was that with a closely fitting jacket, the membrane tubing tolerated, under dynamic conditions, relatively high absolute pressures (studied up to the limit of the fitting used, -700 psi). Since significant pressures were involved for the regenerant flow (up to 300 psi for 5 mL/min) with this design, the actual pressure differential experienced by the membrane tubing was substantially less than the absolute pressure. In the second design, the jacket was concentric with the FFH (Figure 2b). The FFH was made first by tying one end of the filament filled tubing to a suitable support rod, coiling it around the support, and then trying the other end to the support. The whole assembly was then put in boiling water for 30 min to 1h. The thermal treatment set the filament in shape. Ties and support rods were next removed. The coils expanded slightly upon removal of support, the extent of this expansion decreased with decreasing coil diameter and increasing filament diameter. For most of our experiments, support rods 1.5 mm (chromatography tubing) and 0.6 mm (guitar string) in diameter were used; these led to helix diameters in the actual operating devices of about 4 and 2 mm,

respectively. Helices with much larger diameters did not retain their shapes very well and had to be studied by the first design. For operation, the FFH was enclosed in a suitable jacket with inlet/outlet T’s depicted in Figure lb. Tygon or glass tubing was used for the jacket. A closely fitting jacket made efficient use of the regenerant flow; care was taken however to ensure that the FFH did not contact the jacket walls and was stretched very slightly to prevent succeeding turns of the coil from touching each other. In both designs, during coiling, the minimum possible pitch (with turns touching each other) was used. During operation, the regenerant solution was pumped countercurrent to the flow in the FFH. For most of our experiments with design (b), the regenerant solution was alllowed to flow by gravity and, when unimpeded, resulted in a maximum flow rate of 15 mL/min. With design (a), the maximum flow rate of the pump (5.75 mL/min) was used.

RESULTS AND DISCUSSION Regenerant Penetration. At low regenerant concentrations, the Donnan potential effectively prevents the penetration of the regenerant anion through the cation exchange membrane into the inner stream. However, when the concentration of the solution to be ion exchanged is high, regenerant concentration also needs to be relatively high to prevent the mass transport from the outer wall of the membrane to be rate limiting. Under such conditions, the ionic concentration differential between the two sides of the membrane is sufficient to overcome the Donnan potential and penetration of the regenerant anion into the inner flowstream is significant. We compared a number of regenerant solutions with respect to penetration into a water flowstream; the results are reported in Table I. The high penetrations of H3P04and HOAc are clearly due to the efficient transport of the un-ionized uncharged species, not subject to the Donnan barrier. Perchloric and sulfuric acids were studied further, a t higher and more practical concentrations, up to 75 mN; the results are shown in Figure 3. Cation transport through perfluorosulfonate membranes has been studied (24),but anion transport has not been addressed. Clearly, the relationship between penetration rate and regenerant concentration is nonlinear. Likewise, no simple relationship between penetration and osmolality was observed. In our system, the penetration rate was found to be limited by sorption rate at the outer wall or rate of transport through the wall, because under conditions of significant penetration (>lo mN acid), the effluent conductance (directly related to the permeate concentration) was linearly related to the reciprocal of the internal flow rate (data not shown). Helfferich (25) describes the equation governing the equilibrium concentration of a sorbed ion in an ion exchanger bearing fixed charges of the same sign as m / m = (mR/2zm) (a*bP) - mR/2zm (1)

+

where ni and m are the molality of the ion of charge magnitude z in ion exchanger and solution, respectively, mR is the molality of ion exchange groups in ion exchanger, a is the ratio of the activity coefficient of the ion in solution to that in exchanger, b is the ratio of activity of water in exchanger to that in

ANALYTICAL CHEMISTRY, VOL. 56, NO. 1, JANUARY 1984

and malachite green) that bind very strongly to Nafion, were injected into linear and FFH devices. The device was then taken apart and short longitudinal sections, representing equal portions of the tube perimeter, were cut from the membrane tube. The dye was extracted from each section by strongly acidic methanol. Spectrophotometric determination of the dye indicated uniform deposition across the perimeter. Similar experiments with axial cross sections of the membrane tube as a function of the distance from the inlet indicated that the extent of deposition decreased exponentially with increasing distance from the inlet, except in the immediate vicinity of the inlet where the deposition was much higher than that predicted by the exponential relationship. For the three linear annular flow systems (filament radius ri = 0.26,0.28, and 0.32 mm, respectively) the mean annular gap, r, - ri, was calculated from

?

0

40

80

Concentration (mN)

Figure 3. Effect of regenerant concentratlon on penetration. Conditions are the same as those given in Table I.

solution, and p is the ratio of partial molar volume of the electrolyte to partial molar volume of water. In our systems, a, b, m, p , and z may all be regarded as constants as a first approximation. Assuming that the permeate concentration (directly proportional to the observed conductance) is linearly related to m, eq 1 reduces to a much simpler form

aA2

+ @A= m2

(2)

where A is the observed conductance. In Figure 3, the symbols represent the experimental points (molarity rather than molality has been used, the difference is not significant at these concentrations) while the solid lines depict the best fit curve for eq 2 obtained by second-order regression analysis (26). Concentrations lower than 10 mN were omitted from data analysis due to the significant background conductance of the water employed. The fit was good for both HC104 and H2S04; the respective best fit equations were m2 = (3.07 X 10-3)A2 17.16A - 84.58 (R2 = 0.995) and m2 = (4.30 X 10-2)A2+ 53.24A - 166.70 (R2 = 0.996) where m is in units of milliequivalents per liter and A is in microsiemens. The appearance of the constants in the best fit equations is primarily due to the significant conductance of the water used. If lower concentration data are excluded, the magnitude of the constant terfn is reduced drastically. The reduced penetration of H2S04 compared to HC104is probably related to differences in charge and hardness (lipophilicity) of the two anions, aside from the incompleteness of the second dissociation step in the case of H2S04. Note that unlike H3P04or HOAc, incomplete dissociation of HzS04 leaves HS04- which is still subject to the Donnan barrier. On the basis of the above results, H2S04 solutions were used exclusively as regenerant. It should be pointed out that if an electrolyte solution constitutes the inner flowstream, observed penetration appears to be somewhat smaller but occurs nevertheless. The interpretation of the experimental data is however more complex and was not pursued. T h e A n n u l a r Geometry. The regularity of the annular geometry (i.e., does the filament touch the walls of the tube?) is subject to question. Photomicrography of the linear devices or the FFH under operating conditions did not reveal any such irregularity. Inasmuch as the flowing fluid essentially inflates the membrane to create a passage for itself, this is to be expected. However, to determine the nature of mass transfer to the walls of the tube, solutions of cationic dyes (crystal violet

+

99

where V is the dead volume of the device as obtained from residence time measurements (see ref 20 for details). At a flow rate of 2 mL/min, the mean annular gaps were found to be 0.130,0.108,and 0.064 mm, respectively, for the three systems. The outer radii, r,, were then calculated to be 0.390, 0.388, and 0.384 mm, respectively, indicating absence of any large changes in the diameter of the membrane tube between the three systems. However, these values are significantly higher than the mean radius of the hollow tube, determined in an analogous fashion to be 0.350 mm. The annular gap as determined above decreases slightly with decreasing flow rate. Also, the increase in the pressure drop with flow rate is less than linear, suggesting increases in the annular gap with increasing pressure; albeit the effect is small. For the helical devices, the annular gaps, determined as above, were found to be slightly smaller than the corresponding linear devices. The pressure drops at a given flow rate were always larger for the helical devices compared to their linear analogues. For a hollow tube, this increase in pressure drop upon introducing a helical configuration is well-known (10, 11). Mass T r a n s f e r to t h e Wall: Linear Devices. To avoid saturation of the exchange sites, dilute (1 mM) solutions of KNOBand NEt4C104were used as solutions to be ion exchanged. These cations were chosen to explore the effects of varying ionic size and thus the diffusion coefficient. Five millimolar H2S04was used as regenerant, at an adequate flow rate, to minimize regenerant penetration. The fraction exchanged, f, was calculated from f = ( Y - X)/(Z- X) (4) where X is the conductance of the original solution, Y is the conductance of the solution after passage through the device under test, and 2 is the conductance of the solution after passage through a conventional packed column exchanger. The accuracy of this calculation is limited by the pressureinduced water loss from the inside to the outside of the membrane and regenerant penetration, both of which tend to exaggerate the actual efficiency. The results are reported in Table 11. For the NEt4+ case, the initially stable conductance reading decreases eventually to a lower value, due to wall saturation arising from the very slow transport of this ion through the wall. The values reported in Table I1 are the initial values. Some values off, especially those at the lowest flow rate, were calculated to be higher than 1 (up to 1.04) due to regenerant penetration but are reported as unity for consistency. Improvement of efficiency with decreasing annulus width is evident from Table I1 and is observed for both K+ and NEt4+. In quantitatively assessing the increase in efficiency, it is convenient to relate the improvement of performance in terms of decreased device length necessary for a given application. As long as the rate of mass transfer to

100

ANALYTICAL CHEMISTRY, VOL. 56, NO. 1, JANUARY 1984

Table 11. Efficiency of Mass Transfer to the Wall: Fraction Ion Exchangeda flow rate, mL/min device type hollow annular linear, 0.52 mm filament annular linear, 0.56 mm filament annular linear, 0.66 mm filament a

solution

0.5

1.0

2.0

4.0

KNO, NEt4C10, KNO, NEt,ClO, KNO , NEt,ClO, KNO, NEt,ClO,

0.94 0.70 0.94 0.93 1 0.97 1 1

0.64 0.57 0.83 0.77 0.97 0.91 1 0.97

0.54 0.36 0.72 0.58 0.90 0.75 0.98 0.76

0.38 0.29 0.54 0.39 0.67 0.46 0.75 0.56

Tubing length 50 cm in all cases.

the wall is the limiting factor in controlling the observed extent of exchange and irregularities of transport at the inlet/outlet of the device are ignored, an exponential relationship between the device length and the observed extent of exchange may be assumed as a first approximation (analogous to the attenuation of light as a function of path length), such that In (1 - f ) = kL (5)

l 0 O f A

= =

p/

0

where L is the device length and k is a proportionality constant. Thus, a convenient way to consider the increase in efficiency in terms of length is to consider the efficiency factor E

E = In (1- f ? / l n (1- f )

(6)

where f is the extent of exchange observed with the hollow tube and f'is the extent observed with a more efficient device of the same length a t the same flow rate. Limiting our attention to the data for the highest flow rate (4 mL/min) only, the efficiency factors for the 0.52,0.56, and 0.66 mm filament filled systems are 1.6, 2.3, and 2.9 for the K+ case and 1.4, 1.8, and 2.4 for the Wt4+ case, respectively. Values of E calculated with high values o f f are likely to be less reliable due to inaccuracies involved in evaluating f. As discussed in the foregoing section, changes in the annular gap occur for a given system as a function of flow rate. This makes it difficult to carry out theoretical interpretation of the experimental data. Plotting off vs. F for the data in Table I1 (and other data at intermediate flow rates not presented here) show that f is a single valued monotonic function of F for all the systems. The Gormley-Kennedy equation is an infinite series; a valid three term approximation is

+

+

1 - f = 0.8191e-3.657p 0.0975e-22.3p 0.0325e-57p

6.5 40

2.0 1 2 0.65 I

0.2 Dia. ( c m . )

0.4

I

I

(7)

where 1.1 = rDL/F, these terms having been defined previously. While the data for the hollow tube fit reasonably well to eq 7, the calculated values of D from the best fit equations are much larger (by as much as a factor of 4) than the known diffusion coefficients of K+ and NEt4+. Several factors may be responsible for the observed enhancement of transport: pump pulsations, turbulence a t inlet/outlet, turbulence induced by the surface roughness of the tubing, and the finite length necessary for the development of laminar flow. For the filament filled systems, these factors contribute increasingly to the departure from theoretical models, especially with very small annular gaps. Even though the work of Possanzini et al. (17) for annular systems suggests that the GormleyKennedy equation should be applicable to such systems (with the modification that j t should be multiplied by a factor determined by the geometry of the annulus), our data for the annular systems do not fit well to eq 7 . Helical Devices. For a given filament, the annular gap may be considered a constant. At constant rates of flow, the Dean no., Re(d,/dc)l/z, will vary inversely as the square root of the coil diameter, dclIz. Since the Reynold's no., Re, is a constant, d,, which, in the annular case, is the hydraulic radius

6.5 40

2 0 1.2 0.65 0.4 I

0.2 I

Dia. (cm.)

I

Figure 5. Exchange efficiency in helical devices as a function of coil diameter: f mM NEt4CIO4,flow rates as in Figure 4. Device length was 35 cm and filament diameter 0.66 mm.

rather than the tube radius, is &o a constant. Plots of fraction exchanged vs. l/d,l12 are shown in Figures 4 and 5 for K+ and NEt4+, respectively, for 0.66 mm diameter filament filled helices. To obtain a large spread in the value o f f , device lengths chosen for the KNOBand the NEt4C104study were 25 and 35 cm, respectively. If the data for the linear devices are excluded, the plots appear to be monotonous single valued functions with the mass transfer efficiency to the wall being

ANALYTICAL CHEMISTRY, VOL. 50, NO. 1, JANUARY 1984

acutely dependent on the coil diameter. For K+, the efficiency factor E may be calculated from the data presented in Figure 4 to be 3.1 in going from a linear to helical (d, = 0.2 cm) 0.66 mm diameter filament filled device (curve E, flow rate 4 mL/min). Comparison with data presented in Table I1 for the hollow tube (and recognizing the differences in device lengths employed), the efficiency fador for the 0.2 cm diameter helix is then calculated to be 9.5 relative to the hollow tube. The analogous value for NEt4+ at the same flow rate is 8.3 for the 0.66 mm diameter filament filled helix (d, = 0.2 cm) relative to the hollow tube. Experiments with hollow helices as a function of coil diameter show the expected increase in efficiency of mass transfer with decreasing coil radius. However, the effect is substantially lower than that represented by the data for the annular helical case presented above. Within the annular helical geometry, the efficiency was observed to increase with decreasing annular gap (increasing filament diameter), at any given coil diameter. The effect of pitch (defined as the distance between successive turns of the coil divided by the circumference of the coil) at a given coil diameter is increasing efficiency with decreasing pitch, short of the coil turns touching each other during operation. It has been argued that the effect of moderate values of pitch is the same as that due to a small change in the effective coil diameter (IO). Mass Transfer through the Wall. As was noted in the introduction, the outstanding advantage of Nafion perfluorosulfonate membrane is small resistance to cation transport. Consider the limiting case wherein mass transfer to and from the wall are infinitely efficient; the resistance to mass transfer through the wall is the limting factor. This situation plays a significant role in continuous ion exchange of solutions of relatively high concentration such as the eluents used in nonsuppressed ion chromatography (20). For illustrative purposes, consider the ion exchange of a C molar solution of NaX where transport of Na+ through the wall is the only limiting factor. An adequate regenerant flow and concentration is assumed and the rejuvenation of exchange sites by the fast diffusing proton is not considered to be rate limiting. The selectivity coefficient, S, of Nafion for Na+ has been measured by Yeager and Steck (27) to be 1.22; where S is defined as

where Xiis the fraction of ion exchange sites occupied by ion i and [i] represents the molarity of i in solution. (Siincreases at very low values of Xiand a more rigorous analysis will need to take this into account.) Recognizing that the concentration of ion exchange sites is fixed for a given membrane material, eq 8 may be rewritten as SNa+

= MNa+[H'l/((P/Ew) - M N ~ + ) [ N ~ + I (9)

where M iis the concentration of ion i in the exchanger phase in mol/cm3, EW is the equivalent weight of the exchanger, and p is its density; p/EW represents total exchange site concentration in mol/cm3. Rearrangement of eq 9 yields a more convenient form: MNat

= S(p/EW)[Na+]/([H+]

+ S[Na+])

(10)

Consider now an infinitesimally small segment of the device of length dL cm, such that by passage through this section, dC molar Na+ is exchanged at the wall for H+.We assume C >> dC, such that the entire length dL of the segment may be considered to be in equilibrium with C molar Na+. We assume, for simplicity, that the concentration of Na+ outside the membrane (and therefore at the outer surface of the membrane) is essentially zero. If the thickness of the membrane, t , is small, the transport flux of Na+, Q, through the

101

membrane is given by Fick's first law (28)

Q = MNa+Dm/t

(11)

where Q is in mol cm-2 s-l, t is in cm, and D, is the diffusion coefficient of Na+ in the Nafion membrane (measured to be 9.44 X lo-' cm2/s, ref 29). The flux through the membrane for the segment is Q' = 2ariMNa+D,dL/t

(12)

where 2?rridLis the inner surface area of the segment. At steady state with a solution flow rate of F mL/s, the decrease in concentration of Na+ by ion exchange at the wall, -dC, is -dC = lOOOQ'/F

(13)

and the effluent concentration C' is given by

C'= C - dC

(14)

Using a new value of C equal to C' and iterating calculations through eq 10-14, we can evaluate the unexchanged concentration after any number of such segments. However, the change in [H+] needs to be taken into account. From charge balance requirements, at any point in the system [Na+] + [H+] = [OH-]

+ [X-1

(15)

Upon substitution of [Na+] = C where C is the current unexchanged concentration and [X-1 = C,(K/(K + [H+I) where C, is the original concentration of Na+ and K is the dissociation constant of HX, eq 15 yields ["I3

+ (C + K)[H+]' + (KC - K , - KC,)[H+] -

KK, = 0 (16)

Equation 16 may be solved by the Newton-Raphson method (29) at each iteration step and the value fed back in eq 10 for the subsequent cycle. In these calculations, we have assumed t = 0.0075 cm, r = 0.035 cm, and p = 1.98 g/cm3 (the dependence of p on present and past hydration history of the membrane as well as the nature of its ionic content as detailed in ref 30 has been ignored). The numerical solution stabilized at dL = 0.01 cm and thus no further reduction in dL was attempted. Note that for strong acids, eq 16 simplifies to [H+] = C, - C

(17)

and an analytical solution is possible. Equations 10-13 combine to

-dC =

[H+]

+ SC

e a

where 8 = 20007rriD,Sp/(FtEW). Substituting eq 17 into eq 18 and integrating

C = C, exp[-((S - l ) / f

+ W/C))1

(19)

which can be solved numerically. Figure 6 shows the effect of the dissociation constant of HX upon the fraction exchanged. The variation of pH with f is the same as that during the titration of NaX with a strong acid. Consequently,the more facile exchange with decreasing acidity of HX is notable and is particularly important if quantitative or near-quantitative exchange is to be accomplished. Figure 7 shows the effect of varying the concentration of the solution to be ion exchanged (salt of a strong acid). At high concentrations, mass transfer to the wall will clearly be the limiting factor. Figure 8 shows the effect of varying the flow rate of the influent solution (strong acid salt) upon the overall exchange rate. Note that variation of D,, ri, t , or EW lead to similar results.

102

ANALYTICAL CHEMISTRY, VOL. 56, NO. 1, JANUARY 1984

Table 111. Effect of Regenerant Concentration on Ion Exchange Efficiencya regenerant (H,SO,) concn, mM 12.5 25 50

o.2i I 0

4

12

8

16

20

conductance, bS, at the indicated flow rate (mL/min) 0.5

1.0

2.0

0.62 0.70 0.99 2.80

1.77 0.88

7.72 3.05 2.30 2.39

4.0

8.25 6.45 0.73 4.85 100 1.07 4.83 Influent solution 2.4 mM Na,CO, t 3 mM NaHCO, ; tube length 50 cm; filament diameter 0.66 mm; coil diameter 4 mm; design b; jacket i.d. l cm; regenerant flow rate 5 mL/min; direct conductance of this solution is 20.8 p S ; completely exchanged (packed column) conductance is 0.690 US.

1, em

Figure 6. Effect of the dissociation constant of HX on the wall transport Ilmted ion exchange of NaX: (A) pK = 8, (B) pK = 5, (C) pK = 2, (D) strong acid. Influent concentration was 0.01 M; flow rate was 2 mL/min.

L. cm Flgure 7. Effect of the influent concentration on the wall transport limited ion exchange of NaX (HX Is strong acid): (curves A-F) 0.001, 0.005, 0.01, 0.05, 0.1, and 0.5 M influent NaX. Flow rate was 2 mL/min.

L,cm Flgwe 8. Effect of flow rate on the wall transport limited ion exchange of NaX (HX is strong acid): (curves A-D) 0.5, 1.0, 2.0, and 4.0 mL/min. Influent concentration was 0.01 M.

Mass Transport from the Wall. Adequate flow rate and concentration of regenerant are necessary to prevent mass transfer from the outer wall of the membrane from becoming a limiting factor. The maximum permissible degree of regenerant penetration, which must be decided for a specific application, sets an upper limit on the regenerant concentration. Other than the convenience of gravity flow and practicality of maintaining the flow rate, there are no restrictions on increasing the flow rate. The regenerant solution may also be recycled through a large bed exchanger. The geometry of the jacket is obviously a very important factor in determining how efficiently the regenerant flow is utilized. When the influent ionic concentration is high, mass transfer

from the outer wall may become a significant factor. Data for overall exchange observed with a solution containing 3 mM NaHCO, + 2.4 mM Na2C03are presented in Table I11 as a function of regenerant concentration. These data represent an interplay of two opposing factors on the effluent conductance; regenerant penetration and mass transfer inefficiency at the outer wall. Thus, the data in the second column, where the mass transfer efficiency is not limiting, shows increasing conductance with increasing regenerant concentration; this is due to regenerant penetration. In contrast, the data in the last column show the opposite trend; here mass transfer efficiency is the limiting factor. Note also for carbonate containing solution, quantitative cation exchange through the devices often leads to an effluent conductance significantly lower than that obtained with a packed bed exchanger. This is due to facile loss of C02 through the membrane. Since the loss must obviously be much greater with increasing distance from the inlet (as the amount of free H2C03increases), no simple method has proved feasible for calculating the fraction exchanged. Overall Exchange Rate. Since the rate of mass transport from the wall can be prevented from limiting the overall exchange rate by proper choice of regenerant concentration and flow rate, effects due to mass transport from the wall may be disregarded. In general, for most solutions to be ion exchanged, the observed exchange rate will be dependent on mass transfer through the wall for the initial part of the device. This will continue until the concentration drops to a level where the flux brought to the wall falls below the amount that can be transported through the wall. Past this point, the wall will behave as a perfect sink and the exchange rate will be controlled by the rate of mass transport to the wall. Due to the complexity of the hydrodynamics involved, we are presently unable to formulate an analytical expression for the rate of mass transport during helical flow through an annulus. However, as a first approximation, an exponential depletion with length may be assumed under conditions of mass transport to the wall being the limiting factor, for a given flow field. Thus, we may assume that a fraction, k,of the influent flux is capable of reaching the wall during passage through a segment of length dL. The calculations for f as performed in the section on mass transport through the wall then need to be modified only in that -dC = kCdL (20) which is eq 5 in differential form. At each iteration step, either eq 18 or 20 is used depending on which produces the limiting value. The net effect of decreasing k for a given set of wall transport parameters is then manifested in divergence from the wall transport limited exchange curve at lower and lower values of L. Note that k depends on the assumed value of dL,the maximum permissible value being l/dL. However,

103

Anal. Chem. 1984. 56,103-105

for realistic values of the influent concentration, the transfer through the wall is likely to become rate limiting long before this maximum value can be attained.

CONCLUSIONS Although we have focused our attention in this article on the use of the FFH as an efficient continuous cation exchanger, mass transfer efficiency and low dead volume considerations are of paramount importance in many other systems, including dialyzers in artificial kidney machines. Detailed theoretical considerations on the hydrodynamics of helical flow through an annulus are clearly necessary. We believe that such a flow configuration will be of utility in a large number of membrane-based separation systems. The computational approach presented here can be extended to other systems. For example, in designing dialyzers with porous membranes, the density of pore distribution is analogous to exchange site density of an ionomeric membrane, while the selectivity coefficient will be controlled by the pore size distribution and the size of the molecule to be dialyzed. Additionally, in open tubular capillary LC, the generation of active sites in the walls of such small capillaries is a significant task. It is considerably easier to produce such sites on a filament. A thermally shrinkable jacket could then be shrunk down on the filament, yielding very small annular gaps. With appropriate membrane material, an FFH could be used for efficient mixing in of a reagent without large band dispersions. We hope that the considerations outlined in this paper will promote the use of annular and annular helical flow in analytical chemistry. ACKNOWLEDGMENT Discussions with Daniel P. Y. Chang, Department of Civil Engineering, University of California at Davis, were invaluable during all phases of this work. The author is indebted to Donald Chinn, Lubbock High School, for carrying out the bulk of the experimental work. Wescan Instruments is thanked for the gift of the conductivity detector used in this work. Registry No. K, 7440-09-7;NEt4+,66-40-0;H3P04,7664-38-2; HOAc, 64-19-7; HC104,7601-90-3;H2S04,7664-93-9.

LITERATURE CITED (1) Braman, R . S.; Shelley, T. J.; McClenny, W. A. Anal. Chem. 1982, 5 4 , 358-364. (2) Ferm, M. Atmos. Environ. 1979, 73,1385-1393. (3) Gormley, P. 0.; Kennedy, M. Proc. R . I r . Acad., Sect. A 1949, 52A, 163-169. (4) D'Ottavlo, T.; Garber, R.; Tanner, R. L.; Newman, L. Atmos. Environ. 1981, 75, 197-203. ( 5 ) Snyder, L. R.; Kirkland, J. J. "Introduction to Modern Liquid Chromatography", 2nd ed.; Wiley: New York, 1982; p 87. (6) Stevens, T. S.;Davis, J. C.; Small, H. Anal. Chem. 1981, 53, 1488- 1492. (7) Klrkland, J. J., E.I. du Pont de Nemours and Co., personal communlcatlon, Sept 1981. (8) Dean, W. R. Philos. Mag. 1927, 4 , 207-223. (9) Dean, W. R. Phllos. Mag. 1928, 5 , 673-695. (10) Truesdell, L. C., Jr.; Adler, R. J. AIChE J. 1970, 16, 1010-1015. (11) Austln, L. R.; Seader, J. D. AIChE J. 1973, 79, 85-93. (12) Janssen, L. M. Chem. Eng. Sci. 1976, 37,215-218. (13) Deelder, R. S.; Kroll, M. G. F.; Beeren, A. J. B.; Van den Berg, J. H. M. J . Chromatogr. 1978, 749, 669-682. (14) Trivedl, R. N.; Vasudeva, K. Chem. Eng. Sci. 1974, 2 9 , 2291-2295. (15) Trivedi, R. N.; Vasudeva, K. Chem. Eng. Sci. 1975, 30, 317-325. (16) Lundberg, R. E.; Reynolds, W. C.; Kays, W. M. NASA Technical Note D-1972, Aug 1963. (17) Possanzlni, M.; Febo, A,; Libertl, A. Atmos. Environ ., in press. (18) Craig, L. C.; Chen, H. C.; Taylor, W. I. J . Macromoi. Sc;., Chem. 1969, A 3 , 133-149. (19) Stevens, T. S.; Jewett, G. L.; Bredeweg, R. A. Anal. Chem. 1982, 54, 1206-1208. (20) Dasgupta, P. K. Anal. Chem., followlng paper in this issue. (21) Esayian, M., E. I . du Pont de Nemours and Co., personal communication, 1982, 1983. (22) Elsenberg, A., Yeager, H. L., Eds. ACS Symp. Ser. 1982, No. 180. (23) Cox, J. A.; Litwinski, G. R. Anal. Chem. 1983, 55, 1840-1642. (24) Yeager, H. L. ACS Symp. Ser. 1982, No. 180, 41-64. (25) Helfferlch, F. "Ion Exchange"; McGraw-Hill: New York, 1962; p 143. (26) Poole, L.; Borchers, M. "Some Common BASIC Programs", 3rd ed.; Osborne/McGraw-Hill: Berkeley, 1979. (27) Yeager, H. L.; Steck, A. Anal. Chem. 1979, 57, 862-885. (28) Stumm, W.; Morgan, J. J. "Aquatlc Chemistry", 2nd ed.; Wiley: New York, 1981; p 212. (29) Eberhart, J. G.; Sweet, T. R. J. Chem. Educ. 1966, 3 7 , 422-425. (30) E. I . du Pont de Nemours and Co., product information literature, Naflon perfluorosulfonic acid products, 1976.

RECEIVED for review September 15, 1983. Accepted October 7, 1983. This work was supported partially by the Water Resources Center, Texas Tech University, and by the NIH Biomedical Research Support Program at TTU. D.C. was supported by a minority high school student research internship program.

Annular Helical Suppressor for Ion Chromatography Purnendu K. Dasgupta Department of Chemistry, Texas Tech University, Box 4260, Lubbock, Texas 79409

A nylon monofilament filled Nafion perfluorosulfonate catlon exchanger membrane tubing functlons as an efflclent suppressor of low dlsperslon and dead volume for anion chromatography.

Since its introduction in 1975 (I),ion chromatography (IC) with eluent conductance suppression and conductometric detection has established itself as a singularly important and powerful tool, especially in anion analysis. The facility of such analysis as demonstrated by commercially available instrumentation is, at least in part, responsible for resurgent interest in ionic analysis by chromatography. The modern practice of ion chromatography includes alternative detection methods 0003-2700/84/0356-0103$01 SO/O

(2-11). Nonsuppressed single column IC with conductometric detection has also been shown to be useful (12-16). The advantages and disadvantages of nonsuppressed vs. suppressed IC have been discussed in the literature (12, 15,17). In essence, the advantages of suppressed IC include wide dynamic range (large column exchange capacity, low probability of column overloading), rapid equilibration, high sensitivity (good base line stability and low drift), and a wide choice of eluents. It suffers from (a) the need to regenerate the suppressor column, (b) variable retention of weak acids by the suppressor column as a function of its degree of exhaustion, and (c) band broadening induced by the suppressor. With the introduction of the continuously regenerate&hollow fiber suppressor (18) which is always in the same state of regeneration, problems (a) and (b) have been effectively solved. The rather large 0 1983 Amerlcan Chemical Society